proof-aware-crypto-tooling-.../docs/cockpit.md

89 lines
5.1 KiB
Markdown
Raw Normal View History

warden cockpit: local read-only custody UI for the human operator Operator-ordered milestone, built under an explicit no-risk constraint while the paper application is in flight: strictly ADDITIVE (one new module + one CLI subcommand; zero changes to wallet semantics, the public site, or anything the paper describes), LOCAL-ONLY (never deployed; binds 127.0.0.1), and READ-ONLY in v1 (human approve/deny is deliberately deferred — that is a custody-semantics change belonging to a separately reviewed milestone). New: src/pacta/walletui.py — four views over an existing wallet: - Posture: latch state, ledger head with FULL hash-chain re-verification, pinned quorum members, spending policy verbatim. - Signature queue: airgap outbox/inbox, observed never operated. - Incidents & refusals: verbatim records, newest first. - Receipt inspector: paste attestation + receipt + log key; verdict, per-signature results, and diagnostics come verbatim from transparency.verify_receipt — the deployed verifier itself, reused not reimplemented. Design law, stated in the module and enforced by tests: THE COCKPIT RENDERS EVIDENCE, IT NEVER ASSERTS IT. Every panel recomputes at request time and carries a provenance line (function + timestamp); anything unverifiable renders a loud FAILED-TO-VERIFY panel — no cached green, no neutral gray. Panels state what they do NOT prove (e.g. binary pinning vs source-to-binary correspondence). CLI: pacta wallet cockpit --wallet DIR [--host --port]. Tests (10 new, suite 118 -> 128 green): recomputed-evidence rendering, CHAIN BROKEN on tampered ledger, loud failure on collector error, latched-state rendering, airgap queue states, verbatim incident/refusal rendering, fail-closed inspector on garbage, real-evidence roundtrip (skips if example evidence absent), full-sweep BYTE-LEVEL read-only guarantee (wallet dir hash-identical after all GETs + a POST), and fail-fast on non-wallet dirs. Rendering additionally verified visually against a live demo wallet (posture, quorum table, incidents). Docs: docs/cockpit.md; WALLET.md section; ESTATE.md + llms.txt updated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-19 16:02:17 +00:00
# The custody cockpit — a read-only surface for the human operator
`pacta wallet cockpit --wallet <dir>` serves a local web UI
(default `http://127.0.0.1:8471`) over an existing warden wallet.
warden has always been agent-native (MCP) and CLI-native; the cockpit is
the third surface — for the human who ultimately answers for the money.
## The design law
**The cockpit renders evidence; it never asserts it.** Every panel is
recomputed at request time by the same functions the wallet itself uses
(`Wallet.posture()`, `Wallet.verify_ledger()`, directory listings,
`transparency.verify_receipt`), and every panel carries a provenance
line naming the function and the timestamp. Anything that cannot be
recomputed renders as a loud red FAILED-TO-VERIFY panel. There is no
cached green and no neutral gray — a cockpit that shows unverified green
lights would be the anti-warden.
## The UX law (the design law's twin)
**The cockpit never leaves a human in the dark.** A person who has never
heard of warden must be able to read every screen. Concretely, every
page is built from the same anatomy, top to bottom:
1. **Verdict in words** — e.g. CUSTODY HEALTHY / CUSTODY FROZEN
(LATCHED) / CUSTODY EVIDENCE BROKEN — before any evidence, with one
sentence saying what that means and what to do.
2. **A plain-language lead** stating what the page shows and what it
cannot do.
3. **Panels that explain themselves**: each opens with a plain sentence,
carries a "How to read this panel" expander interpreting every column
and every pill, and links each jargon term to the glossary via a
small `?`.
4. **Explained empty states** — an empty list says what empty means and
whether it is good news (for incidents, it is).
5. **The provenance line** — the dashed footer naming the exact function
and timestamp that recomputed the panel.
The `/guide` view is the manual: what warden is, how to read any page,
the color code, a five-minute tour, a glossary of every term (capsule,
member, pinning, evidence grades R0R5, ledger, latch, incident, refusal
receipt, air-gap, attestation/receipt, provenance, DEMO), and an honest
"what this cockpit cannot tell you" section. Navigation tabs state the
question each view answers. This contract is enforced by tests
(`test_guide_view_explains_every_term`,
`test_every_view_carries_lead_nav_and_explainers`,
`test_empty_states_are_explained`).
warden cockpit: local read-only custody UI for the human operator Operator-ordered milestone, built under an explicit no-risk constraint while the paper application is in flight: strictly ADDITIVE (one new module + one CLI subcommand; zero changes to wallet semantics, the public site, or anything the paper describes), LOCAL-ONLY (never deployed; binds 127.0.0.1), and READ-ONLY in v1 (human approve/deny is deliberately deferred — that is a custody-semantics change belonging to a separately reviewed milestone). New: src/pacta/walletui.py — four views over an existing wallet: - Posture: latch state, ledger head with FULL hash-chain re-verification, pinned quorum members, spending policy verbatim. - Signature queue: airgap outbox/inbox, observed never operated. - Incidents & refusals: verbatim records, newest first. - Receipt inspector: paste attestation + receipt + log key; verdict, per-signature results, and diagnostics come verbatim from transparency.verify_receipt — the deployed verifier itself, reused not reimplemented. Design law, stated in the module and enforced by tests: THE COCKPIT RENDERS EVIDENCE, IT NEVER ASSERTS IT. Every panel recomputes at request time and carries a provenance line (function + timestamp); anything unverifiable renders a loud FAILED-TO-VERIFY panel — no cached green, no neutral gray. Panels state what they do NOT prove (e.g. binary pinning vs source-to-binary correspondence). CLI: pacta wallet cockpit --wallet DIR [--host --port]. Tests (10 new, suite 118 -> 128 green): recomputed-evidence rendering, CHAIN BROKEN on tampered ledger, loud failure on collector error, latched-state rendering, airgap queue states, verbatim incident/refusal rendering, fail-closed inspector on garbage, real-evidence roundtrip (skips if example evidence absent), full-sweep BYTE-LEVEL read-only guarantee (wallet dir hash-identical after all GETs + a POST), and fail-fast on non-wallet dirs. Rendering additionally verified visually against a live demo wallet (posture, quorum table, incidents). Docs: docs/cockpit.md; WALLET.md section; ESTATE.md + llms.txt updated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-19 16:02:17 +00:00
## The read-only guarantee
The cockpit cannot approve, sign, unlatch, or modify custody state. It
calls only read paths; the one POST route (the receipt inspector) parses
submitted artifacts in memory and throwaway temp files, never near the
wallet directory. `tests/test_walletui.py` asserts this at the byte
level: a full request sweep, POST included, leaves every file in the
wallet directory hash-identical. Human approve/deny is deliberately NOT
here — that would be a custody-semantics change, which belongs to a
separate, explicitly reviewed milestone.
## The six views
warden cockpit: local read-only custody UI for the human operator Operator-ordered milestone, built under an explicit no-risk constraint while the paper application is in flight: strictly ADDITIVE (one new module + one CLI subcommand; zero changes to wallet semantics, the public site, or anything the paper describes), LOCAL-ONLY (never deployed; binds 127.0.0.1), and READ-ONLY in v1 (human approve/deny is deliberately deferred — that is a custody-semantics change belonging to a separately reviewed milestone). New: src/pacta/walletui.py — four views over an existing wallet: - Posture: latch state, ledger head with FULL hash-chain re-verification, pinned quorum members, spending policy verbatim. - Signature queue: airgap outbox/inbox, observed never operated. - Incidents & refusals: verbatim records, newest first. - Receipt inspector: paste attestation + receipt + log key; verdict, per-signature results, and diagnostics come verbatim from transparency.verify_receipt — the deployed verifier itself, reused not reimplemented. Design law, stated in the module and enforced by tests: THE COCKPIT RENDERS EVIDENCE, IT NEVER ASSERTS IT. Every panel recomputes at request time and carries a provenance line (function + timestamp); anything unverifiable renders a loud FAILED-TO-VERIFY panel — no cached green, no neutral gray. Panels state what they do NOT prove (e.g. binary pinning vs source-to-binary correspondence). CLI: pacta wallet cockpit --wallet DIR [--host --port]. Tests (10 new, suite 118 -> 128 green): recomputed-evidence rendering, CHAIN BROKEN on tampered ledger, loud failure on collector error, latched-state rendering, airgap queue states, verbatim incident/refusal rendering, fail-closed inspector on garbage, real-evidence roundtrip (skips if example evidence absent), full-sweep BYTE-LEVEL read-only guarantee (wallet dir hash-identical after all GETs + a POST), and fail-fast on non-wallet dirs. Rendering additionally verified visually against a live demo wallet (posture, quorum table, incidents). Docs: docs/cockpit.md; WALLET.md section; ESTATE.md + llms.txt updated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-19 16:02:17 +00:00
| view | answers | recomputed by |
warden cockpit: local read-only custody UI for the human operator Operator-ordered milestone, built under an explicit no-risk constraint while the paper application is in flight: strictly ADDITIVE (one new module + one CLI subcommand; zero changes to wallet semantics, the public site, or anything the paper describes), LOCAL-ONLY (never deployed; binds 127.0.0.1), and READ-ONLY in v1 (human approve/deny is deliberately deferred — that is a custody-semantics change belonging to a separately reviewed milestone). New: src/pacta/walletui.py — four views over an existing wallet: - Posture: latch state, ledger head with FULL hash-chain re-verification, pinned quorum members, spending policy verbatim. - Signature queue: airgap outbox/inbox, observed never operated. - Incidents & refusals: verbatim records, newest first. - Receipt inspector: paste attestation + receipt + log key; verdict, per-signature results, and diagnostics come verbatim from transparency.verify_receipt — the deployed verifier itself, reused not reimplemented. Design law, stated in the module and enforced by tests: THE COCKPIT RENDERS EVIDENCE, IT NEVER ASSERTS IT. Every panel recomputes at request time and carries a provenance line (function + timestamp); anything unverifiable renders a loud FAILED-TO-VERIFY panel — no cached green, no neutral gray. Panels state what they do NOT prove (e.g. binary pinning vs source-to-binary correspondence). CLI: pacta wallet cockpit --wallet DIR [--host --port]. Tests (10 new, suite 118 -> 128 green): recomputed-evidence rendering, CHAIN BROKEN on tampered ledger, loud failure on collector error, latched-state rendering, airgap queue states, verbatim incident/refusal rendering, fail-closed inspector on garbage, real-evidence roundtrip (skips if example evidence absent), full-sweep BYTE-LEVEL read-only guarantee (wallet dir hash-identical after all GETs + a POST), and fail-fast on non-wallet dirs. Rendering additionally verified visually against a live demo wallet (posture, quorum table, incidents). Docs: docs/cockpit.md; WALLET.md section; ESTATE.md + llms.txt updated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-19 16:02:17 +00:00
|---|---|---|
| **Posture** (`/`) | *Is custody healthy right now?* Verdict banner, then: custody latch, ledger with full hash-chain re-verification, the pinned quorum members (backend, component, evidence grade, source commit, binary fingerprint), signing rules verbatim, incident/refusal counts | `Wallet.posture()` / `Wallet.verify_ledger()` |
| **Queue** (`/queue`) | *What awaits the offline signer?* Parked air-gap signing requests (outbox) and whether the device has answered (inbox) — observed, never operated | airgap outbox/inbox listing |
| **Incidents** (`/incidents`) | *What has ever gone wrong?* Incident records and signed refusal receipts, verbatim, newest first — with the page explaining why empty is the good state | `incidents/*.json`, `receipts/*.json` |
| **Inspect** (`/inspect`) | *Can I check a receipt myself?* Paste an attestation + transparency receipt + log public key; the verdict, per-signature results, and diagnostics come verbatim from the deployed verifier | `pacta.transparency.verify_receipt` |
| **Estate map** (`/estate`) | *Where does this wallet sit in the whole endeavour?* Every repo, service, mirror, loop — with RUNTIME on every entity (always-on / on-demand / not-running / static) | rendering of ESTATE.md's model (drift-guarded by test) |
| **Guide** (`/guide`) | *What does any of this mean?* The manual: plain-language explanations, color code, tour, full glossary, honest limits — static, no live data | — |
warden cockpit: local read-only custody UI for the human operator Operator-ordered milestone, built under an explicit no-risk constraint while the paper application is in flight: strictly ADDITIVE (one new module + one CLI subcommand; zero changes to wallet semantics, the public site, or anything the paper describes), LOCAL-ONLY (never deployed; binds 127.0.0.1), and READ-ONLY in v1 (human approve/deny is deliberately deferred — that is a custody-semantics change belonging to a separately reviewed milestone). New: src/pacta/walletui.py — four views over an existing wallet: - Posture: latch state, ledger head with FULL hash-chain re-verification, pinned quorum members, spending policy verbatim. - Signature queue: airgap outbox/inbox, observed never operated. - Incidents & refusals: verbatim records, newest first. - Receipt inspector: paste attestation + receipt + log key; verdict, per-signature results, and diagnostics come verbatim from transparency.verify_receipt — the deployed verifier itself, reused not reimplemented. Design law, stated in the module and enforced by tests: THE COCKPIT RENDERS EVIDENCE, IT NEVER ASSERTS IT. Every panel recomputes at request time and carries a provenance line (function + timestamp); anything unverifiable renders a loud FAILED-TO-VERIFY panel — no cached green, no neutral gray. Panels state what they do NOT prove (e.g. binary pinning vs source-to-binary correspondence). CLI: pacta wallet cockpit --wallet DIR [--host --port]. Tests (10 new, suite 118 -> 128 green): recomputed-evidence rendering, CHAIN BROKEN on tampered ledger, loud failure on collector error, latched-state rendering, airgap queue states, verbatim incident/refusal rendering, fail-closed inspector on garbage, real-evidence roundtrip (skips if example evidence absent), full-sweep BYTE-LEVEL read-only guarantee (wallet dir hash-identical after all GETs + a POST), and fail-fast on non-wallet dirs. Rendering additionally verified visually against a live demo wallet (posture, quorum table, incidents). Docs: docs/cockpit.md; WALLET.md section; ESTATE.md + llms.txt updated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-19 16:02:17 +00:00
Every panel also states what it does **not** prove (e.g. the quorum
table says binary hashes are pinned but source-to-binary correspondence
is out of scope until reproducible builds).
## Serving
```bash
pacta wallet cockpit --demo # no wallet yet? throwaway
# DEMO wallet, custody-inert
warden cockpit: local read-only custody UI for the human operator Operator-ordered milestone, built under an explicit no-risk constraint while the paper application is in flight: strictly ADDITIVE (one new module + one CLI subcommand; zero changes to wallet semantics, the public site, or anything the paper describes), LOCAL-ONLY (never deployed; binds 127.0.0.1), and READ-ONLY in v1 (human approve/deny is deliberately deferred — that is a custody-semantics change belonging to a separately reviewed milestone). New: src/pacta/walletui.py — four views over an existing wallet: - Posture: latch state, ledger head with FULL hash-chain re-verification, pinned quorum members, spending policy verbatim. - Signature queue: airgap outbox/inbox, observed never operated. - Incidents & refusals: verbatim records, newest first. - Receipt inspector: paste attestation + receipt + log key; verdict, per-signature results, and diagnostics come verbatim from transparency.verify_receipt — the deployed verifier itself, reused not reimplemented. Design law, stated in the module and enforced by tests: THE COCKPIT RENDERS EVIDENCE, IT NEVER ASSERTS IT. Every panel recomputes at request time and carries a provenance line (function + timestamp); anything unverifiable renders a loud FAILED-TO-VERIFY panel — no cached green, no neutral gray. Panels state what they do NOT prove (e.g. binary pinning vs source-to-binary correspondence). CLI: pacta wallet cockpit --wallet DIR [--host --port]. Tests (10 new, suite 118 -> 128 green): recomputed-evidence rendering, CHAIN BROKEN on tampered ledger, loud failure on collector error, latched-state rendering, airgap queue states, verbatim incident/refusal rendering, fail-closed inspector on garbage, real-evidence roundtrip (skips if example evidence absent), full-sweep BYTE-LEVEL read-only guarantee (wallet dir hash-identical after all GETs + a POST), and fail-fast on non-wallet dirs. Rendering additionally verified visually against a live demo wallet (posture, quorum table, incidents). Docs: docs/cockpit.md; WALLET.md section; ESTATE.md + llms.txt updated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-19 16:02:17 +00:00
pacta wallet cockpit --wallet ~/my-wallet # 127.0.0.1:8471
pacta wallet cockpit --wallet ~/my-wallet --port 9000
```
(Uninstalled, from the repo root:
`PYTHONPATH=src:provider/src python3 -m pacta wallet cockpit --demo`.)
warden cockpit: local read-only custody UI for the human operator Operator-ordered milestone, built under an explicit no-risk constraint while the paper application is in flight: strictly ADDITIVE (one new module + one CLI subcommand; zero changes to wallet semantics, the public site, or anything the paper describes), LOCAL-ONLY (never deployed; binds 127.0.0.1), and READ-ONLY in v1 (human approve/deny is deliberately deferred — that is a custody-semantics change belonging to a separately reviewed milestone). New: src/pacta/walletui.py — four views over an existing wallet: - Posture: latch state, ledger head with FULL hash-chain re-verification, pinned quorum members, spending policy verbatim. - Signature queue: airgap outbox/inbox, observed never operated. - Incidents & refusals: verbatim records, newest first. - Receipt inspector: paste attestation + receipt + log key; verdict, per-signature results, and diagnostics come verbatim from transparency.verify_receipt — the deployed verifier itself, reused not reimplemented. Design law, stated in the module and enforced by tests: THE COCKPIT RENDERS EVIDENCE, IT NEVER ASSERTS IT. Every panel recomputes at request time and carries a provenance line (function + timestamp); anything unverifiable renders a loud FAILED-TO-VERIFY panel — no cached green, no neutral gray. Panels state what they do NOT prove (e.g. binary pinning vs source-to-binary correspondence). CLI: pacta wallet cockpit --wallet DIR [--host --port]. Tests (10 new, suite 118 -> 128 green): recomputed-evidence rendering, CHAIN BROKEN on tampered ledger, loud failure on collector error, latched-state rendering, airgap queue states, verbatim incident/refusal rendering, fail-closed inspector on garbage, real-evidence roundtrip (skips if example evidence absent), full-sweep BYTE-LEVEL read-only guarantee (wallet dir hash-identical after all GETs + a POST), and fail-fast on non-wallet dirs. Rendering additionally verified visually against a live demo wallet (posture, quorum table, incidents). Docs: docs/cockpit.md; WALLET.md section; ESTATE.md + llms.txt updated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-19 16:02:17 +00:00
The server binds localhost by default and is not meant to be exposed;
there is no authentication because there is nothing to operate.